Explore · 每周精选
发现最受关注的研究论文,追踪研究趋势,订阅感兴趣的期刊与关键词。
Haozhe Zhao, Shuzheng Si, Zhenhailong Wang +6 authors
Automated systems for generating scientific figures face limitations in handling diverse figure types and conditions, prompting the development of multi-agent frameworks that generalize across different input scenarios and produce editable output formats.
50 篇论文 · 按点赞排序
Mind Lab, Song Cao, Vic Cao +51 authors
Parameter-efficient fine-tuning can function as a compact substrate for persistent personal models by enabling small trainable adapters to store instance-specific behaviors on top of strong foundation models.
Jianuo Huang, Yaojie Zhang, Qituan Zhang +3 authors
Domino is a speculative decoding framework that improves LLM inference speed by decoupling causal dependency modeling from autoregressive drafting through a parallel backbone and lightweight causal refinement head, achieving significant speedups in both end-to-end execution and throughput.
Aditi, Niket Agarwal, Arslan Ali +288 authors
Cosmos 3 is an omnimodal world model that processes and generates multiple data types through a unified mixture-of-transformers architecture, achieving state-of-the-art performance in various understanding and generation tasks.
Tianyi Zhou, Dongrui Liu, Leitao Yuan +2 authors
Person-grounded AI skills are automatically distilled from heterogeneous traces into inspectable, correctable packages that capture both capabilities and behavioral patterns.
Zhifei Xie, Zihang Liu, Ze An +8 authors
A unified streaming audio model is developed that combines offline task execution with real-time audio instruction following through an end-to-end framework supporting multiple audio interaction capabilities.
Alireza Salemi, Chang Zeng, Atharva Nijasure +4 authors
GrepSeek enables efficient search agent training through direct corpus interaction using shell commands and a two-stage approach combining a cold-start dataset and group relative policy optimization.
Maksim Savkin, Mikhail Goncharov, Alexander Gambashidze +7 authors
Compact task-specialized language models demonstrate superior performance in multi-hop reasoning and faithfulness compared to larger general-purpose models through a novel training pipeline and structured reasoning traces.
Liliana Hotsko, Yinxi Li, Yuntian Deng +1 authors
Code2LoRA is a hypernetwork framework that generates repository-specific LoRA adapters for code language models, supporting both static and evolving codebases with efficient parameter-efficient fine-tuning.
Tomer Keren, Nitay Calderon, Asaf Yehudai +3 authors
Automated benchmark generation method creates challenging tasks with broader tool-use coverage by evolving tool sequences through adaptive contrastive n-gram modeling and iterative difficulty refinement.
Lorenz K. Muller, Philippe Bich, Chiara Boretti +3 authors
KVarN is a calibration-free KV-cache quantizer that uses Hadamard rotation and dual-scaling variance normalization to reduce error accumulation during autoregressive decoding in large language models.
Daniil Plyusov, Alexey Gorbatovski, Alexey Malakhov +4 authors
Trust-Region behavior Blending improves on-policy distillation by replacing early poor-quality student rollouts with teacher-like behavior within a KL trust region during warmup.
Ruiqi Li, Yu Zhang, Changhao Pan +3 authors
A zero-shot text-to-speech system called SwanVoice is presented that addresses expressive long-form multi-speaker dialogue synthesis by combining VAE, flow-matching DiT, and diffusion post-training techniques.
Haoxiang Zhang, Qixin Xu, Zhuofeng Li +4 authors
Observation masking in long-horizon search agents shows variable accuracy gains depending on the interaction between retriever capability and model capacity, following an asymmetric inverted-U pattern.
Yuqing Wang, Zhijie Lin, Ceyuan Yang +10 authors
Representation Forcing enables unified multimodal models to perform both perception and generation tasks end-to-end without relying on external latent spaces, matching state-of-the-art performance in image generation while improving understanding capabilities.
Pengcheng Jiang, Zhiyi Shi, Kelly Hong +5 authors
A 20B search agent trained with reinforcement learning within a stateful search framework demonstrates superior retrieval performance across multiple domains by separating semantic decision-making from environmental bookkeeping.
Nahyun Lee, Dongkeun Yoon, Guijin Son +12 authors
Korean web-browsing agent benchmark K-BrowseComp evaluates frontier LLMs' capabilities with 400 problems, showing significant performance gaps compared to English benchmarks and highlighting the need for more robust Korean AI development.
Marko Kojic, Ivan Bondyrev, Aral de Moor +6 authors
Mellum 2 is an open-weight 12B-parameter Mixture-of-Experts language model with 2.5B active parameters per token, specialized in software engineering tasks and optimized for inference efficiency on commodity GPUs.
Jiaming Wang, Ziteng Feng, Jiangtao Wu +8 authors
Deep-research agents can be audited using a claim-centric framework that identifies error spans in their reasoning trajectories, improving reliability assessment beyond just final answer evaluation.
Yuval Golbari, Navve Wasserman, Matias Cosarinsky +5 authors
BrainCause framework uses generative and brain models to identify valid neural representations through causal testing, demonstrating that activation alone is insufficient for confirming concept representation.
Sihyun Yu, Nanye Ma, Pinzhi Huang +8 authors
Current multimodal large language models struggle with visual state tracking in videos, performing poorly even when human-level capabilities are required, and existing agentic approaches do not effectively address these limitations.
Woojung Song, Nalim Kim, Sangjun Song +3 authors
Role-playing language agents require dynamic character development that evolves through narratives, necessitating benchmarks that evaluate psychological trajectory alignment rather than static factual recall, with ArcANE demonstrating superior performance when character arc information is conditioned into models.
Xingrun Xing, Haoqing Wang, Boyan Gao +2 authors
Trust Region On-Policy Distillation (TrOPD) improves reliable token-level supervision in large language model distillation by using trust regions, outlier estimation, and off-policy guidance to address instability issues under distribution mismatch.
Soyeong Jeong, Jinheon Baek, Minki Kang +1 authors
TIDE is a template-guided iterative framework for discovering hidden problems from context, using iterative discovery and thought templates to improve problem identification and resolution across document and code environments.
Xiangtao Kong, Jixin Zhao, Lingchen Sun +2 authors
Generative multimodal foundation models are used to create high-quality training data for image restoration, improving model generalization across diverse real-world scenarios.
Jiayu Liu, Cheng Qian, Zhenhailong Wang +8 authors
AdaPlanBench presents a dynamic interactive benchmark for evaluating LLM agents' ability to adaptively plan under progressively revealed world and user constraints through multi-turn interactions.
Xuekang Wang, Zhuoyuan Hao, Shuo Hou +3 authors
CHERRL is a controlled environment for studying reward hacking in rubric-based reinforcement learning with LLM judges, enabling detection and analysis of subtle bias exploitation patterns.
Zekun Qi, Xuchuan Chen, Dairu Liu +10 authors
Humanoid-GPT is a GPT-style Transformer with causal attention trained on a billion-scale motion corpus that achieves zero-shot generalization to unseen motions and control tasks through scalable pre-training on diverse motion data.
Nianyi Lin, Jiajie Zhang, Lei Hou +1 authors
LongTraceRL addresses long-context reasoning challenges in large language models through tiered distractor construction and rubric reward design for improved reasoning quality.
Ruiqi Wang, Qimin Chen, Daniel Ritchie +4 authors
Function2Scene generates 3D indoor layouts from functional descriptions by parsing user needs and applying design constraints through an iterative refinement process combining geometric analysis, language modeling, and visual assessment.
北京市昌平区探索星信息技术及软件开发工作室
京ICP备2026059466号